583 ·GitHub
What is this?
Intrinsic Core is an open, local runtime, SDK and hardware-agnostic real-time control framework for industrial robotics. It provides foundational building blocks for AI-enabled robot applications, from simulation and 6-DoF perception down to hardware execution.
Intrinsic Core is an open, local runtime, SDK and hardware-agnostic real-time control framework for industrial robotics. It provides foundational building blocks for AI-enabled robot applications, from simulation and 6-DoF perception down to hardware execution.Is it a runtime, a framework, or an SDK?All three, delivered as one stack: intrinsicruntime is a pre-configured local execution engine, ICON is the real-time motion and hardware coordination engine, and intrinsicsdk is the toolkit for building reusable skills, custom hardware assets and execution nodes on top.
All three, delivered as one stack: intrinsicruntime is a pre-configured local execution engine, ICON is the real-time motion and hardware coordination engine, and intrinsicsdk is the toolkit for building reusable skills, custom hardware assets and execution nodes on top.How do I get it running?You need Ubuntu 24.04 LTS or 26.04 LTS (22.04 is supported too) with the ROS 2 Lyrical Luth distribution; the Getting Started guide covers full installation and setup. The runtime itself is packaged as a native k3s containerized environment.
You need Ubuntu 24.04 LTS or 26.04 LTS (22.04 is supported too) with the ROS 2 Lyrical Luth distribution; the Getting Started guide covers full installation and setup. The runtime itself is packaged as a native k3s containerized environment.Do I need to know ROS 2 to use it?It builds on ROS 2 rather than replacing it: the runtime gives you a stable, deterministic execution layer on top of ROS 2 so you skip manual low-level system plumbing, while the interoperability bridges modern C++/Bazel development with the broader ROS ecosystem.
It builds on ROS 2 rather than replacing it: the runtime gives you a stable, deterministic execution layer on top of ROS 2 so you skip manual low-level system plumbing, while the interoperability bridges modern C++/Bazel development with the broader ROS ecosystem.Is this an officially supported Google product?No — the README explicitly disclaims that it is not an officially supported Google product. It is maintained as an open project with a developer community and forum, and contributions require signing the Google Contributor License Agreement.
No — the README explicitly disclaims that it is not an officially supported Google product. It is maintained as an open project with a developer community and forum, and contributions require signing the Google Contributor License Agreement.Can I use it in a commercial robotics project?Yes — it is licensed under Apache 2.0, which permits commercial use, modification and redistribution. Keep in mind that 'Intrinsic' and 'Intrinsic Core' are trademarks of Intrinsic Innovation LLC with their own usage guidelines.
Yes — it is licensed under Apache 2.0, which permits commercial use, modification and redistribution. Keep in mind that 'Intrinsic' and 'Intrinsic Core' are trademarks of Intrinsic Innovation LLC with their own usage guidelines.Does the AI perception part need an NVIDIA GPU?Out of the box, perception embeds NVIDIA FoundationPose for 6-DoF pose estimation of 3D parts, and the inference engine is built to run such vision models accelerated directly on edge hardware. Exact GPU requirements are not stated, but the stack is NVIDIA-centric.
Out of the box, perception embeds NVIDIA FoundationPose for 6-DoF pose estimation of 3D parts, and the inference engine is built to run such vision models accelerated directly on edge hardware. Exact GPU requirements are not stated, but the stack is NVIDIA-centric.What can it do out of the box?Grasp planning, AI perception, a native digital twin and collision-free motion planning come pre-built and ready to use. Through the SDK you extend this with reusable skills, custom hardware assets and execution nodes that integrate with the digital twin and runtime.
Grasp planning, AI perception, a native digital twin and collision-free motion planning come pre-built and ready to use. Through the SDK you extend this with reusable skills, custom hardware assets and execution nodes that integrate with the digital twin and runtime.What is the Open Machine Tending Solution (OMTS)?An open-source reference design built on Intrinsic Core for a real machine shop use case. It ships pre-configured assets, skills and best practices as a functional machine tending template, offering a zero-refactor path from prototype to deployment.
An open-source reference design built on Intrinsic Core for a real machine shop use case. It ships pre-configured assets, skills and best practices as a functional machine tending template, offering a zero-refactor path from prototype to deployment.Can I swap robot arms and grippers without rewriting drivers?That is the design goal: a standardized Hardware Abstraction Layer (HAL) enables sensor-driven control across arms, grippers and fieldbus I/O without hardware lock-in, while the intrinsichardware module supplies drivers, manifests and integrations for industrial peripherals.
That is the design goal: a standardized Hardware Abstraction Layer (HAL) enables sensor-driven control across arms, grippers and fieldbus I/O without hardware lock-in, while the intrinsichardware module supplies drivers, manifests and integrations for industrial peripherals.Is the control loop deterministic enough for real hardware?Yes: ICON combines kinematics solvers and trajectory interpolation with a deterministic control loop that can switch controllers within a single cycle based on live sensor feedback. The companion malloc-guard library even prevents dynamic heap allocations in real-time threads.
Yes: ICON combines kinematics solvers and trajectory interpolation with a deterministic control loop that can switch controllers within a single cycle based on live sensor feedback. The companion malloc-guard library even prevents dynamic heap allocations in real-time threads.What are similar projects?FluxVLA, InstinctFlash, cs249r_book, LightNav-0, GR00T-WholeBodyControl, Zetta-Embodiment, ZDTaichu5.0-9B, quackd
FluxVLA, InstinctFlash, cs249r_book, LightNav-0, GR00T-WholeBodyControl, Zetta-Embodiment, ZDTaichu5.0-9B, quackdWhich topics is it in?ros2, physical-ai, manufacturing, embodied-ai, edge-ai, robotics, simulation, gpu-acceleration, hardware, 3d, real-time, kubernetes
ros2, physical-ai, manufacturing, embodied-ai, edge-ai, robotics, simulation, gpu-acceleration, hardware, 3d, real-time, kubernetes